Slide 1

Helge Meinhard CERN CERN Openlab Summer Student Lectures 2014 From Detectors to Publications Slide 2 1.Brief introduction to CERN 2.Overview of process from detectors to results 3.Reconstruction 4.Simulation 5.Physics Analysis 6.Computing 7.Summary Outline Acknowledgement: Contents partly based on 3-part lecture given in 2013 summer student lecture programme Thanks to Jamie Boyd et al. Slide 3 Slide 4 CERN International organisation close to Geneva, straddling Swiss-French border, founded 1954 Facilities for fundamental research in particle physics 21 member states, 1 B CHF budget 3360 staff, fellows, students, apprentices, 11000 users Helge Meinhard - From detectors to publications4 Science for peace 1954: 12 Member States Members: Austria, Belgium, Bulgaria, Czech republic, Denmark, Finland, France, Germany, Greece, Hungary, Israel, Italy, Netherlands, Norway, Poland, Portugal, Slovak republic, Spain, Sweden, Switzerland, United Kingdom Candidate for membership: Romania Associate member: Serbia Observers: European Commission, India, Japan, Russia, Turkey, UNESCO, United States of America Numerous non-member states with collaboration agreements Members: Austria, Belgium, Bulgaria, Czech republic, Denmark, Finland, France, Germany, Greece, Hungary, Israel, Italy, Netherlands, Norway, Poland, Portugal, Slovak republic, Spain, Sweden, Switzerland, United Kingdom Candidate for membership: Romania Associate member: Serbia Observers: European Commission, India, Japan, Russia, Turkey, UNESCO, United States of America Numerous non-member states with collaboration agreements 2512 staff members, 540 fellows, 315 students, 24 apprentices 6700 member states, 1800 USA, 900 Russia, 236 Japan, 03-Jul-2014 Slide 5 CERN where the Web was born Helge Meinhard - From detectors to publications5 03-Jul-2014 Slide 6 Slide 7 LHC Large Hadron Collider Proton-proton collider 27 km circumference Started operation in 2010 with 3.5 + 3.5 TeV, continued in 2011, 4 + 4 TeV in 2012 Worlds most powerful particle accelerator Run 1 until early 2013 Now in Long Shutdown 1 machine upgrade Restart early 2015 at 6.5 + 6.5 TeV Helge Meinhard - From detectors to publications7 03-Jul-2014 Slide 8 Four Large Detectors ATLAS, CMS, ALICE, LHCb Some ATLAS facts: 100 million channels 25 m diameter, 46 m length, 7000 tons 3000 scientists (including 1000 grad students) 40 MHz collision rate Run 1: 300 Hz event rate after filtering All LHC experiments: 30000 TB in 2012, 100000 TB in total Helge Meinhard - From detectors to publications8 03-Jul-2014 Slide 9 Helge Meinhard - From detectors to publications9 03-Jul-2014 Slide 10 Results so far Many the most spectacular one being 04 July 2012: Discovery of a Higgs-like particle March 2013: The particle is indeed a Higgs boson 08 Oct 2013 / 10 Dec 2013: Nobel price to Peter Higgs and Fran ois Englert CERN, ATLAS and CMS explicitly mentioned Helge Meinhard - From detectors to publications10 03-Jul-2014 Slide 11 Helge Meinhard - From detectors to publications11 03-Jul-2014 Slide 12 Physics Results ARE NOT Finding a single golden event (or a few of them) Helge Meinhard - From detectors to publications12 03-Jul-2014 Slide 13 Physics Results ARE statistical comparisons of experimental data with theoretical predictions Helge Meinhard - From detectors to publications13 03-Jul-2014 Slide 14 14 Theory... eg. the Standard Model has parameters coupling constants masses predicts: cross sections, branching ratios, lifetimes,... Helge Meinhard - From detectors to publications 03-Jul-2014 Slide 15 Experiment 150 million active elements 20 (40) million bunch crossings per second O(1 PB/s) internal data rate Data reduction: Suppress electronic noise Decide to read out and save event, or throw it away (trigger) Build the event (assemble all data) O(300 Hz) event rate O(500 MB/s) data rate Helge Meinhard - From detectors to publications15 03-Jul-2014 Slide 16 Compare Theory with Experiment Define observables: Physics quantities that are sensitive to variations in theoretical predictions Cross sections (probability of a certain type of interaction to occur) Branching ratios Particle masses Particle lifetimes Calculate statistically these observables over a large number of events recorded Calculate statistically these observables as you expect them from theory Compare the two Fit parameter(s) Check fit quality Helge Meinhard - From detectors to publications16 03-Jul-2014 Slide 17 Analysis Flow Detector & Trigger Reconstruction Physics Analysis Compare theory and experiment Reconstruct event candidates 1.Select events 2.Calculate observables 3.Cut on observables Simulated data Reconstruction Physics Analysis 17 Slide 18 Detector (and Trigger) (1) Proton-proton collisions create numerous secondary particles at interaction point These secondary particles must be detected Helge Meinhard - From detectors to publications18 03-Jul-2014 Slide 19 Detector (and Trigger) (2) Particle detectors generally consist of two major elements: Tracking detectors in the centre: measure precisely tracks of secondary particles Calorimeters surrounding: Let particles dump all their energy and measure the energy deposited A magnetic field in the detector forces charged particles on a helix path Helge Meinhard - From detectors to publications19 03-Jul-2014 Slide 20 Detector (and Trigger) (3) Secondary particles are detected by their ionising effect on the detector material Different techniques: gaseous devices, semiconductor devices, light-emitting devices, Every cell can deliver a single measurement Often 3-dimensional, sometimes 2-dimensional points Detector/trigger/data acquisition system deliver raw data of an event: a collection of such points Helge Meinhard - From detectors to publications20 03-Jul-2014 Slide 21 Helge Meinhard - From detectors to publications21 03-Jul-2014 Slide 22 Reconstruction From points to identified particles: Detector reconstruction Tracking finding path of charged particles through the detector determining momentum, charge, point of closest approach to interaction point Calorimeter reconstruction finding energy deposits in calorimeters from charged and neutral particles determining energy deposit, location, and direction Combined reconstruction Electron/photon identification Muon identification Jet finding Helge Meinhard - From detectors to publications22 03-Jul-2014 Slide 23 Reconstruction: Figures of Merit Efficiency How often do we reconstruct the object correctly? Resolution How accurately do we reconstruct it (momentum / energy, vertex position etc.)? Fake rate How often do we reconstruct an object different from the real one? These figures depend on Detector Reconstruction algorithms, calibration, alignment We want/need High efficiency, good resolution, low fake rate to know the efficiency, resolution, fake rate Robustness against detector problems Fit into computing resource limitations (CPU, memory) Helge Meinhard - From detectors to publications23 03-Jul-2014 Slide 24 Tracking: An Example Where is the 50 GeV track? Helge Meinhard - From detectors to publications24 03-Jul-2014 Slide 25 Tracking Issues Ionisation is a statistical process Ionisation means energy loss: momentum and direction of original particle get (slightly) changed Detector elements not perfectly aligned (LHC) Pile-up: a single collision of two proton bunches results in more than one proton-proton collision Up to around 30 in Run 1, expected > 50 in Run 2 Helge Meinhard - From detectors to publications25 03-Jul-2014 Slide 26 Pile-up Example Helge Meinhard - From detectors to publications26 03-Jul-2014 Slide 27 Z0Z0 q q e+e+ e-e- particle detector element Helge Meinhard - From detectors to publications27 03-Jul-2014 Slide 28 Why Simulation? Compare observables with expectations from theoretical models, which leads to physics results Design detectors Optimise trigger settings Tune analysis selections Estimate background and systematic errors Estimate efficiency, resolution and fake-rate Helge Meinhard - From detectors to publications28 03-Jul-2014 Slide 29 Simulation Steps: 1. Physics Simulate physics interaction at proton-proton collision Input: parameters of physics model Output: events with four-vectors of secondary particles Z0Z0 q q e+e+ e-e- Helge Meinhard - From detectors to publications29 03-Jul-2014 Slide 30 Simulation Steps: 2. Detector Simulate interaction of secondary particles with the detector material Includes ionisation, bending of charged particles in magnetic field, Requires very complete and detailed detector description Based on standard toolkit GEANT 4 Very CPU-intensive (more so than reconstruction) particle detector element Helge Meinhard - From detectors to publications30 03-Jul-2014 Slide 31 Simulation Steps: 3. Electronics Simulate the response of the detector elements to interactions of secondary particles passing through Requires very detailed description of detector electronics Technically often combined with step 2 (detector simulation) GEANT 4 Output is very similar to detector raw data Information of truth kept all the way through Helge Meinhard - From detectors to publications31 03-Jul-2014 Slide 32 125 GeV Helge Meinhard - From detectors to publications32 03-Jul-2014 Slide 33 Types of Physics Analyses Search for new particles / phenomena Usually limited by statistics Negative result (nothing found) sets new exclusion limits Precision measurements Usually limited by systematic uncertainties Serve as consistency check of the underlying theory (mostly the Standard Model) Helge Meinhard - From detectors to publications33 03-Jul-2014 Slide 34 Physics Analysis Workflow (1) Start with the output of reconstruction Apply an event selection based on the reconstructed object quantities Often calculate new information, e.g. masses of combinations of particles Event selection designed to improve the signal to background in your event sample Estimate Efficiency of selection (& uncertainty) Background after selection (& uncertainty) Can use simulation for these but have to use data-driven techniques to understand the uncertainties Make final plot Comparing data to theory Correcting for efficiency and background in data Include the statistical and systematic uncertainties Helge Meinhard - From detectors to publications34 03-Jul-2014 Slide 35 Physics Analysis Workflow (2) Output of reconstruction is large (order of tens of PBs); simulation similarly large LHC experiments have defined reduced data sets (both general and specific to certain physics channels) Sometimes, final steps of analysis are based on an even more reduced private data set Many of the analyses use standard packages for their final steps including presentation of plots and histograms, e.g. ROOT Helge Meinhard - From detectors to publications35 03-Jul-2014 Slide 36 Helge Meinhard - From detectors to publications36 03-Jul-2014 Slide 37 The Nature of the Problem Enormous numbers of collisions of proton bunches with each other Data from each collision are small (order 110 MB) Each collision independent of all others No supercomputers needed Most cost-effective solution is standard PC architecture (x86) servers with 2 sockets, SATA drives, Ethernet network Linux (Scientific Linux, RHEL variant) used everywhere Calculations are mostly combinatorics integer (rather than floating-point) intensive Helge Meinhard - From detectors to publications37 03-Jul-2014 Slide 38 The Scale of the Problem 2012: 30000 TB of new data from LHC experiments CD tower of 50 km height, or 1000 years of videos on DVD Requires 250000 fast compute cores, 200000 TB disk space, 207000 TB tape space CERN able to provide some 15% of this capacity Helge Meinhard - From detectors to publications38 03-Jul-2014 Slide 39 A distributed computing infrastructure to provide the production and analysis environments for the LHC experiments Managed and operated by a worldwide collaboration between the experiments and the participating computer centres The resources are distributed for funding and sociological reasons Our task was to make use of the resources available to us no matter where they are located Helge Meinhard - From detectors to publications WLCG what and why? Tier-0 (CERN): Data recording Initial data reconstruction Data distribution Tier-1 (11 centres): Permanent storage Re-processing Analysis Tier-2 (~130 centres): Simulation End-user analysis 39 Slide 40 CERN Computer Centre 15% of LHC compute requirements Services for other (smaller) experiments Computing infrastructure for CERN and the collaborations hosted January 2014: 10500 servers, 76000 disk drives, 52000 tape cartridges Total power/cooling envelope for IT equipment: 3.5 MW Not sufficient for requirements of LHC Run 2 Helge Meinhard - From detectors to publications40 03-Jul-2014 Slide 41 Extension: Wigner data centre in Budapest, Hungary Result of open tender in CERN member states First few 100 machines in production Final envelope: 2.5 MW Helge Meinhard - From detectors to publications41 03-Jul-2014 Slide 42 CERN openlab in a nutshell A science industry partnership to drive R&D and innovation with over a decade of success Evaluate state-of-the-art technologies in a challenging environment and improve them Test in a research environment today what will be used in many business sectors tomorrow Train next generation of engineers/employees Disseminate results and outreach to new audiences Helge Meinhard - From detectors to publications42 03-Jul-2014 Slide 43 Helge Meinhard - From detectors to publications43 03-Jul-2014 Slide 44 Physics results are statistical comparisons of observables with their theoretical predictions Trigger/data acquisition, reconstruction, analysis, and simulation is how we get there Sophisticated algorithms needed for reconstruction and analysis Calibration and alignment must be well controlled Detailed simulation is key to the success The single event is small and simple, but the computing scale is enormous Software and computing must work very well in order to achieve results with the LHC Helge Meinhard - From detectors to publications44 03-Jul-2014 Slide 45 Thank you